The world of quantitative finance continues to evolve at a rapid pace. Even in the last four years of the existence of this site the market for quant jobs has shifted significantly. In this article we outline these shifts. The advice on what is likely to be in demand in the next few years will be applicable both to those still in education as well as those thinking ahead to a career change.
The rate of information diffusion has increased significantly in the previous few years. There are many contributing factors to this trend. The rise of abundant sensor data capture has provided vast troves of data to analyse. Sophisticated open-source software is now competitive with commercial offerings in many arenas. Lower barriers to entry along with leveraged cloud computing resources allow small teams and individuals to compete with larger firms in many areas of technology.
This trend towards increased availability of information and leveraged mechanisms for analysing it shows no sign of deceleration. In addition the rapid changes in the financial technology sector, including the significant compliance and regulatory overhead, have caused substantial shifts in the marketplace for available quantitative finance career opportunities.
Over the last few years the cost of acquiring quality asset pricing data, at least for more liquid exchange-traded assets, has dwindled. The immense competition between data vendors has reduced prices substantially while still maintaining a high degree of quality.
Vendors such as Quandl, QuantQuote, QuantGo as well as all-in-one cloud trading platforms like Quantopian have had a huge impact on the diffusion of available data.
The simultaneous rise of open-source freely-available languages such as Python, including all of its libraries such as NumPy, SciPy, Pandas, Scikit-Learn, HDF5 and TensorFlow have "levelled the playing field" for smaller teams in comparison to the larger firms, certainly in terms of data analysis capabilities.
The 2007/2008 subprime crisis and subsequent regulatory and compliance regime change has dramatically reduced the profitability of much of the derivatives pricing business that was carried out pre-crisis. The emphasis for banks carrying out this work is now heavily on model validation, stress-testing and risk management, particularly in the areas of counterparty credit risk and default.
Alternative "non-traditional" data from many sources are introduced to the market on a daily basis. Previously unattainable sources such as periodic high-resolution global satellite imagery is now available commercially from multiple competing startup providers. Huge troves of agricultural and industrial sensor data (via the "AgTech" and "Internet of Things" sectors) is also available, albeit less easily, driving vast research (both academic and commercial) in analysing this data for informational edges that can be profited from.
It is clear to see that such changes in information availability have had a dramatic effect on the quant finance job market even over the last few years. The hiring emphasis is now on "FinTech", "big data analytics" and "data science", with multiple funds and venture-backed startup data providers vying to hire the best data science talent from top universities and other sectors.
Despite these vast changes the market for traditional deriviatives pricing - particularly in counterparty risk and portfolio risk management - is still very healthy. The need for quantitative skills has never been stronger. Highly stimulating - and lucrative - career options remain both in quant finance and data science.
In the current market and coming years the following sets of skills are likely to be in high demand.
As the world continues its long march towards automation so increases the demand for competent software developers. While the need for software engineers is readily apparent from the techology sector, particularly in venture capital backed technology startups, there is also still strong latent demand in the financial sector.
The decline in traditional lines of business (such as derivatives pricing, see above), along with the heavy regulation and compliance costs of carrying out prop trading, is leading investment banks to look elsewhere for paths to revenue. The rise of the FinTech (Financial Technology) sector means that technologists versed in both high finance and software development are very much in demand.
In particular, the core set of imperative object-oriented languages including C++11, C#, Java, Scala and Python admit plenty of job opportunities, either full-time, contract or freelance.
Perhaps the most advantageous aspect of gaining skills in software development is the implicit hedge provided through the ability to transition across sectors, from quant finance, through tech startups and even academia (many labs are always looking for competent developers to code up newly researched models!). This is a strong consideration if youre thinking about choosing your undergraduate degree topic/major.
While software developers can earn a strong base salary, particularly in certain areas of quant finance or data-driven technology startups, their upside potential is often limited unless there is significant bonus-pool potential or clear equity arrangements with the employing firm. The latter area is extremely complex from a taxation standpoint (not to mention varying cross-jurisdictionally!) and so it will not be discussed in any further depth.
Statistical Machine Learning and Deep Learning
One of the most often-discussed topics on this site is machine learning. It is an extremely useful tool for financial data analysis. It is also currently one of the most attractive career skillsets for quant researchers to possess. Machine learning skills are also highly sought after in other sectors, including insurance, consumer technology startups and agricultural technology (AgTech) companies.
A good understanding of Bayesian inference and machine learning opens up many opportunities in "data science" and roles in "big data". For those with particular aptitudes it also allows a transition between academia and commercial research firms, leading to a highly intellectually stimulating career that is both financially lucrative and extremely rewarding.
Very recently the academic and commercial research community has focussed on Deep Learning, which is an umbrella term encompassing many machine learning methods involving network architectures with many ("deep") layers. This interest has largely been due to the staggering performance of deep architecture models on classification tasks, as well as its highly effective application to search engine results and social networks.
Quant funds have been in the game of machine learning and deep learning for some time. There is a vast literature on applications of "traditional" analysis methods such as linear time series models as well as the more elementary ML methods such as kernel- or tree-based classifiers and regressors.
The literature becomes somewhat more sparse in the application of deep learning architectures and reinforcement learning techniques to quant finance, although a good write-up can be found at Greg Harris' blog. The lack of information can be explained either by the fact that serially-correlated data is harder to analyse with deep learning tools or that such techniques are so effective that quant funds are unwilling to publish their hard-won IP.
Nevertheless there are many firms who will hire entire research groups from some of the top deep learning research departments. The lure of "millions of people" using their work, along with the highly attractive remuneration, is often too tempting to turn down.
It is too early to say at this stage whether deep learning is a passing fad or is actually an area of ML that will be around for some time. The latter is more probable, given its effectiveness on many tasks that were considered impossible, or at the very least, a long way off even less than a year ago.
Our recommendation is to become highly skilled in the foundations of statistical machine learning and then leverage that knowledge to become an expert at deep learning. If such models fall out of favour then it will be straightforward to transition to the next promising area of ML technology.
Stochastic Calculus and Financial Engineering
We have discussed the inappropriateness of gaining a Masters in Financial Engineering solely for the purpose of becoming a quant trader at length before on QuantStart. But that is not to say that they aren't still relevant. In fact the opposite is true.
The increased regulatory and compliance burden placed on banks and asset managers has seen an increase in the need for risk managers and model validation quants. The traditional highly mathematical techniques from stochastic calculus and derivatives modeling are still widely used in these instances.
In addition there are many traditional independent "shops" that still carry out a substantial amount of derivatives trading and/or risk management, particularly in commodities, foreign exchange, fixed income and of course, credit risk.
These skills are best picked up both in a highly mathematical degree such as mathematics or theoretical physics, along with tuition from a top-tier school, predominantly in the US or the UK. The latter requirement is mainly for the "brand value" when applying to recruiters for roles post-MFE.
However, as we have emphasised before on the site, do not be under the illusion that an MFE is the path to riches via a cushy quant portfolio management role. There are vast numbers of students graduating MFE courses from universities with high variability in their quality. It will be hard to stand out in such a crowded space, especially if you do not have additional skills upon which to draw, such as programming expertise or machine learning competence.
With a specialisation in this skillset, be prepared for early career roles in model validation or risk management/oversight, rather than the elusive "junior quant trader" roles often touted by recruiters.
Econometrics, Time Series Analysis and Frequentist Statistics
While sophisticated quant hedge funds and investment banks have largely moved away from traditional time series analysis tools, there are other types of firms that still consider these skills valuable.
In particular, an interesting trend that members of QuantStart are personally familiar with is the rise of the quant trading arm within family office structures.
Entrepreneurs who have managed to have a large (liquid!) "exit" from their businesses are now more often than not forming a family office structure to professionally manage their assets. Upon hiring some of the best asset managers to run these firms the need for diversification in a world of near-zero or negative interest rates, and a likely average real-return on equities of around 3-4%, is prompting asset managers to form a quant trading divisions.
Many such efforts begin with a small team. Their first task is often to accumulate quality data and build trading infrastructure. Traditional tools such as time series analysis and econometrics are encouraged - standard institutional benchmarks are less appropriate in the family office environment. Time frames can be much longer and these techniques remain competitive over these durations.
Hence if job seeking candidates have a good network, possess expertise in time series analysis and traditional time series econometrics, this provides one avenue of oppotunity. However it will require a more sophisticated approach to job hunting than "going through the front door".
Non-Traditional Skills - GIS, Remote Observation, Sensor Fusion
Alternative data sources are now becoming the latest targets for hedge funds who are desperately searching for new sources of alpha. As with traditional fundamental data, government economic reports and asset pricing data, the need for a fund to have access to a data source solely because their key competitors do leads to an inevitable data "arms race".
Venture-backed AgTech and IoT data vendors are founded regularly due to the huge drops in cost for both sensor hardware (such as drones and satellites) along with staggering rises in performance from computer vision software (deep neural network learning architectures). This has had a huge impact on how institutions trade commodities markets.
The rise of these vendors has generated a need for individuals skilled in non-traditional scientific analysis, such as sensor fusion/analytics, remote observation, GIS capability along with more common data science skills such as the aforementioned machine learning and deep learning capability.
Such advances, while highly exciting from an intellectual standpoint, make it challenging for the prospective student or career-changer to know what skills to specialise, or retrain, in.
The best advice is to make sure that your underlying quantitative capability is strong. Extensive skills in mathematics, physics, statistics and/or computer science are always going to be in demand, irrespective of the current focus on particular technologies. With a strong quantitative base the transition into the latest technology of the day is always possible.